Multi-temporal Images Research Articles

Assessment of ensemble learning for object-based land cover mapping using multi-temporal Sentinel-1/2 images

Accurate land cover mapping is challenging in Southeast Asia where cloud coverage is prevalent and landscape is heterogenous. Object-based mapping, multi-temporal images and combined use of optical and microwave data, provide abundant features in spectral, spatial, temporal, geometric and polarimetric dimensions. And random forest is usually employed due to robustness and efficiency in handling high-dimensional and noisy data. This study assesses whether feature selection and ensemble analysis, which are rarely adopted, yield improved result. Recursive feature elimination decreases original 568 features into a subset of 53 features, achieving the optimal combination of features. Ensemble analysis of random forest, support vector machine, and K-nearest neighbors leads to an overall accuracy of 0.816. Comparison experiments demonstrated the merits of the multi-temporal, multi-source approach, feature elimination and ensemble analysis.

EZ-InSAR: An easy-to-use open-source toolbox for mapping ground surface deformation using satellite interferometric synthetic aperture radar

Satellite Interferometric Synthetic Aperture Radar (InSAR) is a space-borne geodetic technique that can map ground displacement at millimetre accuracy. Via the new era for InSAR applications provided by the Copernicus Sentinel-1 SAR satellites, several open-source software packages exist for processing SAR data. These packages enable one to obtain high-quality ground deformation maps, but still require a deep understanding of InSAR theory and the related computational tools, especially when dealing with a large stack of images. Here we present an open-source toolbox, EZ-InSAR (easy-to-use InSAR), for a user-friendly implementation of InSAR displacement time series analysis with multi-temporal SAR images. EZ-InSAR integrates the three most popular and renowned open-source tools (i.e., ISCE, StaMPS, and MintPy), to generate interferograms and displacement time series by using these state-of-art algorithms within a seamless Graphical User Interface. EZ-InSAR reduces the user’s workload by automatically downloading the Sentinel-1 SAR imagery and the digital elevation model data for the user’s area of interest, and by streamlining preparation of input data stacks for the time series InSAR analysis. We illustrate the EZ-InSAR processing capabilities by mapping recent ground deformation at Campi Flegrei (> 100 mm·yr−1) and Long Valley (~ 10 mm·yr−1) calderas with both Persistent Scatterer InSAR and Small-Baseline Subset approaches. We also validate the test results by comparing the InSAR displacements with Global Navigation Satellite System measurements at those volcanoes. Our tests indicate that the EZ-InSAR toolbox provided here can serve as a valuable contribution to the community for ground deformation monitoring and geohazard evaluation, as well as for disseminating bespoke InSAR observations for all.

Impact assessment of severe cyclonic storm Asani on the nesting grounds of Olive Ridley turtle, Rushikulya Estuary and spit in Odisha state, India

The objective of this study is to understand the impact of cyclone Asani on the coastal dynamics and nesting grounds of Olive Ridley sea turtles in the Rushikulya Estuary and spit in Odisha. The novelty of this study is the need to understand the effects of natural disasters on the marine environment and wildlife. This study provides new and valuable data on the effects of specific cyclones on sea turtle populations and highlights the need for increased attention and protection of these important habitats. The period of shoreline change analyzed in this study was before (May 2, 2022) and after (May 17, 2022) the impact of cyclone Asani. Multi-temporal Sentinel 2A images and Digital Shoreline Analysis System were used to perform such an analysis. Field surveys were conducted using a Leica DGPS ArcPad, and group discussions with relevant stakeholders such as the fishing community, government officials, and coastal conservation guards were also undertaken. The findings of this study showed that the cyclone had a significant impact on the spit, dividing it into four smaller pieces and opening three new mouths. The spit experienced erosion and a significant reduction in length, and local fishing community reported that 15–20% of sea turtle eggs were washed away. This study highlights the severe threat to the survival of Olive Ridley sea turtles in Odisha due to changing cyclone patterns and the potential for degradation of the local ecosystem. The results of this study can aid in coastal conservation and management strategies to protect marine turtle populations in the region.

A New Method for False Alarm Suppression in Heterogeneous Change Detection

Heterogeneous change detection has a wide range of applications in many fields. However, to date, many existing problems of heterogeneous change detection, such as false alarm suppression, have not been specifically addressed. In this article, we discuss the problem of false alarm suppression and propose a new method based on the combination of a convolutional neural network (CNN) and graph convolutional network (GCN). This approach employs a two-channel CNN to learn the feature maps of multitemporal images and then calculates difference maps of different scales, which means that both low-level and high-level features contribute equally to the change detection. The GCN, with a newly built convolution kernel (called the partially absorbing random walk convolution kernel), classifies these difference maps to obtain the inter-feature relationships between true targets and false ones, which can be represented by an adjacent matrix. We use pseudo-label samples to train the whole network, which means our method is unsupervised. Our method is verified on two typical data sets. The experimental results indicate the superiority of our method compared to some state-of-the-art approaches, which proves the efficacy of our method in false alarm suppression.

Mapping and assessing spatial extent of floods from multitemporal synthetic aperture radar images: a case study over Adyar watershed, India.

Urban floods are more concerned in recent days due to their substantial effect in loss of human lives and properties. Due to climate change, urban floods are frequently observed in many parts of the world. Flood events in Chennai city are a frequent scenario due to rapid increase in the density of population. Adyar river watershed and surrounding urban cover are focused in the present study. The present study aims at mapping flooded region using Sentinel 1A datasets over Adyar watershed. Series of Sentinel 1A image is collected before, during and after floods for mapping the extent of flood and mapping risk zones in Adyar watershed. Methodologies such as ISODATA Technique, Multi-Temporal Analysis, Thresholding Method, PCA and ICA Analysis and Grey Level Co-Occurrence Matrix are adopted for the extraction of flooded extent from the SAR datasets. Analysis performed over the Adyar watershed provided promising results in the extraction of flooded extent with Thresholding Method and Grey Level Co-Occurrence Matrix being the dominant of all the methods. Though higher accuracy is obtained in the extraction of flooded extent, limitation of layover, foreshortening and shadow is experienced in the built up region for the extraction of flooded pixels.

Fusion of spectral and spatial information for agricultural crop classification in multi-temporal Sentinel images (Case Study: Qorveh County)

امروزه کاربردهای تصاویر ماهواره‌ای، در پایش و مدیریت زمین‌های کشاورزی، رو به گسترش است. با توجه به قدرت تفکیک مکانی، طیفی و زمانی بالای تصاویر سنتینل‌ـ 2، در این مطالعه، در کشاورزی دقیق در شهرستان قروه از این تصاویر استفاده شده است. ابتدا با توجه به تقویم زراعی محصولات متفاوت آن منطقه، تصاویر سری زمانی جمع‌آوری شد. در روش پیشنهادی، نخست، فضای ویژگی طیفی براساس بازتاب طیفی باندها و همچنین شاخص‌های گیاهی، ایجاد شد. ابعاد فضای ویژگی طیفی، با استفاده از روش آنالیز مؤلفه‌های اصلی، کاهش یافت. سپس چهار طبقه‌بندی‌کنندة قدرتمند ماشین‌های بردار پشتیبان، شبکة عصبی پرسپترون چندلایه، نزدیک‌ترین k همسایه و جنگل‌های تصادفی نقشة طبقه‌بندی از اطلاعات طیفی تولید کردند. در ادامه، مکانی با هدف تعیین مرز مزارع، اطلاعات استخراج شد. برای این منظور، از شناسایی لبه‌ها در سری زمانی تصاویر سنتینل‌ـ 2 استفاده شد. در نهایت، نقشة طبقه‌بندی نهایی، با تلفیق اطلاعات مکانی و ادغام نتایج طبقه‌بندی‌کننده‌ها ایجاد شد. نتایج به‌دست‌آمده نشان داد که دقت طبقه‌بندی‌کننده‌های نزدیک‌ترین k همسایه، ماشین‌های بردار پشتیبان، شبکة عصبی پرسپترون چندلایه و جنگل‌های تصادفی روی فضای ویژگی طیفی اولیه، به‌ترتیب 78/77%، 16/79%، 41/76% و 89/76% است. با استفاده از روش پیشنهادی، دقت طبقه‌بندی به 72/94% افزایش پیدا کرد که حاکی از توانایی آن در منطقة مورد مطالعه است.

Monitoring and assessment of urban green space loss and fragmentation using remote sensing data in the four cities of Malawi from 1986 to 2021

Urban green spaces enhance cities' natural beauty, boost biodiversity, and promote the health and happiness of city residents, among other benefits. Despite the benefits of urban green spaces, their spatial configuration and composition are impacted by urban expansion. This study aims to investigate the spatial-temporal fragmentation trends of urban green spaces in four cities in Malawi (Blantyre, Zomba, Lilongwe, and Mzuzu). We, therefore, applied multi-temporal Landsat images to understand trends in urban green spaces from 1986 to 2021. Five landscape metrics, namely: percentage of land area (PLAND), number of patches (NP), patch density, largest patch index (LPI), landscape shape index (LSI), and mean patch size (MPS), were used to quantify the degree of urban green space fragmentation in Fragstats software. The results showed an increase in the NP, PD, and LSI and a decline in the LPI and MPA in the four cities. This is an indication of fragmentation and a reduction in green space. Rapid urbanisation has resulted in the loss and fragmentation of urban green space areas, which has been attributed to the rapid expansion of built-up areas at the expense of green spaces. The present study is the first to incorporate landscape metrics to understand the spatial heterogeneity of the composition and configuration of green spaces in all the cities of Malawi. The findings of this study contribute to an understanding of how urban growth impacts fragmentation and the loss of urban green spaces and offer valuable insights for better spatial planning and management of cities.

Climate Change Impacts on LULC in the Jarmet Wetland and its Surrounding Areas in Western Ethiopia

The concerns over land use and land cover (LULC) change have emerged on the global stage due to realization that changes occurring on the land surface also influence climate, ecosystem and its services. This study aimed to map the temporal dynamic of LULC patterns and LST in the Jarmet wetland in Ethiopia. The dynamics and pattern of changes for a period of 21 years (2000-2021) were analyzed using geospatial techniques. Multi-temporal satellite images from Landsat ETM+ and Landsat-8 OLI sensor data were used to extract land-cover maps. The Land Surface Temperature (LST) trend of the study areas was computed using MODIS satellite imagery (2000-2021). Supervised classification using a Maximum Likelihood Classifier (MLC) was applied to prepare LULC maps of the watershed. The accuracy of the classified map was assessed using high-resolution data, and ground realities have been verified and ascertained through field observations. The results revealed a decreased trend in wetland, forest, shrubland and grassland in the period of 21 years (2000-2021) by -1148.71ha, -1073.26 ha, -1480.1 ha, and -87.73 ha, respectively. On the other hand, farmland and plantation areas followed an increasing trend. LST revealed decreasing trend in terms of mean and minimum with a fraction change of -0.018 and -0.073, whereas the maximum LST value shows an increasing trend with 0.021. The overall accuracy was 84.41%, with Kappa index of 76.13%. The analysis and findings of the study highlight important policy implications for sustainable LULC management in the study area. The study suggests the design and implementation of a guided natural resource policy, stopping the illegal expansion of farmland and educating society about the value of the sustainable management of habitat reserves.

Shoreline change rate estimation: Impact on salt production in Kutubdia Island using multi-temporal satellite data and geo-statistics

The coastline is a natural interface that separates the wet and dry parts of the beach. Shoreline configuration is always influenced by natural and anthropogenic forces. Storm surge is one of the major natural events that has converted the shape of shoreline along the Kutubdia island. With the use of the Digital Shoreline Analysis System (DSAS), the coastal area was evaluated to estimate the change dynamics of Kutubdia Island. This highly active coast with erosion and deposition has inherent complexity in determining the shoreline movement and quantifying the rate of change. The study examined the positional change of the shoreline of the island between 1980 and 2021 including its impacts on salt production using Geographic Information System (GIS) tools and multi-temporal LANDSAT images. To distinguish between water and land features, the LANDSAT images were radiometrically corrected and a spectral index, the Normalized Difference Water Index (NDWI) was used. The shorelines were extracted using a histogram-based Otsu's Binary threshold approach and image-based visual interpretation. The Shoreline Change Envelope (SCE), Linear Regression Rate (LRR), and End Point Rate (EPR) was used to calculate shoreline change rates. EPR and LRR both estimated average change rates of −4.125 m/yr and −4.647 m/yr, respectively. Between 1980 and 2021, an overall land area of 9.2439 sq. km was lost, while there were no significant land depositions. The landward movement was more noticeable on the northern side of the island. The eastern side showed no significant alterations. The short-term (in 2031) and long-term (in 2041) coastline positions have been forecast based on the demarcated shoreline. The results of the forecast also show that the shoreline will move landward but not significantly seaward. Island erosion is particularly dangerous in the northern and southern zones. The result also shows that salt production is decreasing due to island erosion, especially in the south. If the erosion continues at the same rate salt production will decrease almost by 4563.38 tons between 2021 and 2041. The results of this study may help with the proper management and planning for Kutubdia Island. Moreover, this can also be used in statistical modeling, to predict where the shoreline will be in the future.

Impact of urban and peri-urban growth on arable land (1976–2029) in a medium sized city of Shire Indaselassie, North Western Tigray, Ethiopia

Urban and peri-urban areas have been growing rapidly globally due to population increase and other factors. However, poor urban land use planning caused remarkable impacts on the loss of potentially arable land. This study aimed to analyze the spatio-temporal patterns of urban and peri-urban expansion and its impact on arable land in the Shire Indaselassie city, North Western Tigray. Multi-temporal and spectral Landsat images were used as an input and processed using TerrSet, ERDAS imagine 2015, and ArcGIS 10.8.1 softwares. The Cellular Automata-Markov Chain model was used to predict the future urban and peri-urban areas. Pearson correlation coefficient was applied to examine the association among the population, urban, peri-urban growth, and its impact on arable land in Stata/SE v14. The findings revealed that urban and peri-urban lands increased by 7.5 km2 (22.1%) and 1.0 km2 (2.9%), while arable land decreased by − 0.1 km2 (− 0.2%) from 1976 to 2019. It is also predicted to increase by 1.0 km2 (2.80%) and 0.9 km2 (2.70%) until 2029, while arable land is anticipated to shrink by − 1.0 km2 (− 3.0%). The relationship among population growth, urban, and peri-urban land shows positive (r = 0.942, p = 0.058; r = 0.985, p = 0.016). However, arable land has been negatively correlated with population growth, urban, and peri-urban lands (r = − 0.610, p = 0.392). The result of this research is indispensable for urban planners and decision makers to optimize urban and peri-urban expansion related information without jeopardizing the size of arable lands.

Modelling and Analysis of Urban Growth of Idah Metropolis Using GIS Integrated Gradient Technique

This thesis aims to better understand changes in the spatial pattern of urban growth and its impact on landscape configuration by conducting a case study in Idah metropolis. The objectives are as follows: 
 
 To model and analyze multi-temporal images (2001, 2010, and 2021) for determining the urban growth in Idah,
 To determine urban development trends using post-classification comparisons,
 Evaluating the degree of urban growth according to urban land cover change over that period 
 
 Satellite imagery was employed to distinguish and identify different land surface categories. Integrated remote sensing and GIS (Geographic Information System) technique was used to analyse both qualitative and quantitative perspectives regarding the objectives. The results indicate that the urban area of Idah experienced changes in the various classes. The vegetated area experienced a great change of 49.98%. That is a decrease from the previous year of 2001 to 2010. Also, another 48.67% decrease from 2010 to 2021 is due to the developments in the study area. The built-up area equally decreased by 1.33 % between 2010 and 2021. There was an increase of 12.04% from the year 2001 to 2010.Beach has a drastic increase from 28.10% observed between 2001 and 2010 to 44.50% between 2010. The geographic footprint demonstrates that the distribution of the built-up area was dispersed and continues to grow more dispersed. The significant contribution of this study could benefit many aspects, such as comparative studies between cities or continuous studies relevant to urban growth.

Optimal Time Phase Identification for Apple Orchard Land Recognition and Spatial Analysis Using Multitemporal Sentinel-2 Images and Random Forest Classification

The significance of identifying apple orchard land and monitoring its spatial distribution patterns is increasing for precise yield prediction and agricultural sustainable development. This study strived to identify the optimal time phase to efficiently extract apple orchard land and monitor its spatial characteristics based on the random forest (RF) classification method and multitemporal Sentinel-2 images. Firstly, the Normalized Difference Vegetation Index (NDVI), Soil-Adjusted Vegetation Index (SAVI), Ratio Vegetation Index (RVI), and Difference Vegetation Index (DVI) between apple orchard land and other green vegetation (other orchards, forest and grassland) during the growing stage were calculated and compared to identify the optimal time phase for apple orchard land extraction; the RF classifier was then constructed using multifeature variables on Google Earth Engine to efficiently identify apple orchard land, and the support vector machine (SVM) classification results were used as a comparison; GIS spatial analysis, a slope calculation model, and Moran’s I and Getis-Ord GI* analysis were employed to further analyze the spatial patterns of the apple orchard land. The results found the following: (1) April, May, and October were the optimal time phases for apple orchard identification. (2) The RF-based method combining coefficients of indexes, the grayscale co-occurrence matrix, and 70% of the ground reference data can precisely classify apple orchards with an overall accuracy of 90% and a Kappa coefficient of 0.88, increasing by 9.2% and 11.4% compared to those using the SVM. (3) The total area of apple orchard land in the study area was 485.8 km2, which is 0.6% less than the government’s statistical results. More than half (55.7%) of the apple orchard land was distributed on the gentle slope (Grade II, 6–15°) and the flat slope (Grade I, 0–5°); SiKou, Songshan, and Shewopo contained more than 50% of the total orchard land area. (4) The distribution of apple orchard land has a positive spatial autocorrelation (0.309, p = 0.000). High–High cluster types occurred mainly in Sikou (60%), High–Low clusters in Songshan (40%), Low–High clusters in Sikou (47.5%), and Low–Low clusters in Taocun and Tingkou (37.4%). The distribution patterns of cold and hot spots converged with those of the Local Moran Index computation results. The findings of this study can provide theoretical and methodological references for orchard land identification and spatial analysis.

Tropical Dry Forest Dynamics Explained by Topographic and Anthropogenic Factors: A Case Study in Mexico

Tropical dry forest is one of the most threatened ecosystems, and it is disappearing at an alarming rate. Shifting cultivation is commonly cited as a driver of tropical dry forest loss, although it helps to maintain the forest coverage but with less density. We investigated tropical dry forest dynamics and their contributing factors to find out if there is an equilibrium between these two processes. We classified multi-temporal Sentinel-2A images with machine learning algorithms and used a logistic regression model to associate topographic, anthropogenic, and land tenure variables as plausible factors in the dynamics. We carried out an accuracy assessment of the detected changes in loss and gain considering the imbalance in area proportion between the change classes and the persistence classes. We estimated a 1.4% annual loss rate and a 0.7% annual gain rate in tropical dry forest and found that the topographic variable of slope and the anthropogenic variable of distance to roads helped explain the occurrence probability of both tropical forest loss and tropical forest gain. Since the area estimation yielded a wide confidence interval for both tropical forest loss and gain despite the measures that we took to counterbalance the disproportion in areas, we cannot conclude that the loss process was more intense than the gain process, but rather that there was an equilibrium in tropical dry forest dynamics under the influence of shifting cultivation.

Mapping seasonal changes of street greenery using multi-temporal street-view images

Street greenery offers various benefits to urban environments. In regions with climatic variations among seasons, seasonal changes of vegetation may lead to fluctuations in the benefits provided by street greenery. It is vital to monitor and measure the seasonal changes of street greenery. Previous studies have analyzed changes of street greenery, mainly from an aerial view. However, aerial views may not be equivalent to residents' visual experiences. This study aims to quantitatively characterize seasonal differences in street greenery based on multi-temporal street-view images which can simulate pedestrians' view. The Gulou District in Nanjing, China, is selected for a pilot study. We collected multi-temporal street-view images through an online street-view service. Deep learning-based algorithms were used to extract seasonal street greenery from street-view images. The results revealed significant seasonal differences in street greenery in the Gulou District. We classified four street greening patterns, including (1) Deciduous and evergreen mixed pattern; (2) Deciduous-dominant pattern; (3) No-plant pattern; (4) Evergreen-dominant pattern, with distinct seasonal change characteristics. For each pattern, we explored possible adjustments in planting arrangements. Our work is a preliminary attempt, and the proposed framework could assist in future sustainable greening design and planning.

The spatial distribution and expansion of Eucalyptus in its hotspots: Implications on agricultural landscapes

Fast coppicing plantations like Eucalyptus are becoming an ever increasingly important land use system globally, including the Eucalyptus hotspot highlands of Northwestern Ethiopia. However, comprehensive information regarding species composition is essential for proper planning and policy decisions. The current study mapped the spatial distribution of Eucalyptus globulus (hereafter referred to as Eucalyptus) and identified the key push factors for its expansion. The study used a mapping procedure that uses Landsat imagery together with ground truth data based on supervised training of a pixel-by-pixel classification algorithm within image regions to distinguish areas of Eucalyptus plantations from other classes. High-resolution multispectral and multi-temporal remote-sensing images were combined with ground truth data to produce robust features of Eucalyptus plantation distribution maps. Heckman's Two-Stage econometric model was also employed for determining the major driving factors of Eucalyptus expansion. The results of the mapping algorithm were Eucalyptus plantation distribution maps of 30 × 30 m resolution that showed temporal changes from 1999 to 2021. The findings revealed that Eucalyptus coverage increased by 55% during the period from 1999 to 2010 and the change expressively increased to 69% in 2021 with respect to the reference period. The study also found that a number of push factors influenced the size of land planted with Eucalyptus. The developed maps showing the spatial distribution and expansion of Eucalyptus will help policymakers properly manage the ecosystems and agricultural landscapes of Eucalyptus growing areas.

Spatial Pattern and Intensity Mapping of Coseismic Landslides Triggered by the 2022 Luding Earthquake in China

On 5 September 2022, an Mw 6.6 earthquake occurred in Luding County in China, resulting in extensive surface rupture and casualties. Sufficient study on distribution characteristics and susceptibility regionalization of the earthquake-induced disasters (especially coseismic landslides) in the region has great significance to mitigation of seismic hazards. In this study, a complete coseismic landslide inventory, including 6233 landslides with 32.4 km2 in area, was present through multi-temporal satellite images. We explored the distribution and controlling conditions of coseismic landslides induced by the 2022 Luding event from the perspective of epicentral distance. According to the maximum value of landslide area density, the geographical location with the strongest coseismic landslide activity intensity under the influence of seismic energy, the macro-epicenter, was determined, and we found a remarkable relationship with the landslide distribution and macro-epicentral distance, that is, both the landslide area and number density associatively decreased with the increase in macro-epicentral distance. Then, a fast and effective method for coseismic landslide intensity zoning based on the obvious attenuation relationship was proposed, which could provide theoretical reference for susceptibility mapping of coseismic landslides induced by earthquakes in mountainous areas. Additionally, to quantitatively assess the impact of topographic, seismogenic and lithological factors on the spatial pattern of coseismic landslides, the relationships between the occurrences of coseismic landslides and influencing factors, i.e., elevation, slope angle, local relief, aspect, distance to fault and lithology, were examined. This study provides a fresh perspective on intensity zoning of coseismic landslides and has important guiding significance for post-earthquake reconstruction and land use in the disaster area.

Coastal landscape classification using convolutional neural network and remote sensing data in Vietnam

The length of global coastline is about 356 thousand kilometers with various dynamic natural and anthropogenic. Although the number of studies on coastal landscape categorization has been increasing, it is still difficult to distinguish precisely them because the used methods commonly are traditional qualitative ones. With the leverage of remote sensing data and GIS tools, it helps categorize and identify a variety of features on land and water based on multi-source data. The aim of study is using different natural – social profile data obtained from ALOS, NOAA, and multi-temporal Landsat satellite images as input data of the convolutional-neural-network (CvNet) models for coastal landscape classification. Studies used 900 cut-line samples which represent coastal landscapes in Vietnam for training and optimizing CvNet models. As a result, nine coastal landscapes were identified including: deltas, alluvial, mature and young sand dunes, cliff, lagoon, tectonic, karst, and transitional landscapes. Three CvNet models using three different optimizer types classified the landscapes of other 1150 cut-lines in Vietnam with the accuracies about 98% and low loss function value. Excepting dalmatian, karst and delta coastal landscapes, five others distribute heterogeneous along the coasts in Vietnam. Therefore, the evaluation of additional natural components is necessary and CvNet model have ability to update new landscape types in variety of tropical nation as a step toward coastal landscape classification at both national and global scales.

Bishift Networks for Thick Cloud Removal with Multitemporal Remote Sensing Images

Because of the presence of clouds, the available information in optical remote sensing images is greatly reduced. These temporal-based methods are widely used for cloud removal. However, the temporal differences in multitemporal images have consistently been a challenge for these types of methods. Towards this end, a bishift network (BSN) model is proposed to remove thick clouds from optical remote sensing images. As its name implies, BSN is combined of two dependent shifts. Moment matching (MM) and deep style transfer (DST) are the first shift to preliminarily eliminate temporal differences in multitemporal images. In the second shift, an improved shift net is proposed to reconstruct missing information under cloud covers. It introduces multiscale feature connectivity with shift connections and depthwise separable convolution (DSC), which can capture local details and global semantics effectively. Through experiments with Sentinel-2 images, it has been demonstrated that the proposed BSN has great advantages over traditional methods and state-of-the-art methods in cloud removal.

Global and Local Graph-Based Difference Image Enhancement for Change Detection

Change detection (CD) is an important research topic in remote sensing, which has been applied in many fields. In the paper, we focus on the post-processing of difference images (DIs), i.e., how to further improve the quality of a DI after the initial DI is obtained. The importance of DIs for CD problems cannot be overstated, however few methods have been investigated so far for re-processing DIs after their acquisition. In order to improve the DI quality, we propose a global and local graph-based DI-enhancement method (GLGDE) specifically for CD problems; this is a plug-and-play method that can be applied to both homogeneous and heterogeneous CD. GLGDE first segments the multi-temporal images and DIs into superpixels with the same boundaries and then constructs two graphs for the DI with superpixels as vertices: one is the global feature graph that characterizes the association between the similarity relationships of connected vertices in the multi-temporal images and their changing states in a DI, the other is the local spatial graph that exploits the change information and contextual information of the DI. Based on these two graphs, a DI-enhancement model is built, which constrains the enhanced DI to be smooth on both graphs. Therefore, the proposed GLGDE can not only smooth the DI but also correct the it. By solving the minimization model, we can obtain an improved DI. The experimental results and comparisons on different CD tasks with six real datasets demonstrate the effectiveness of the proposed method.

USING MACHINE LEARNING CLASSIFIERS AND REGRESSION MODELS FOR ESTIMATING THE STAND AGES OF FALCATA PLANTATIONS FROM SENTINEL DATA

Abstract. Falcata is a widely planted Industrial Tree Plantation (ITP) species in the Caraga Region, Mindanao, Philippines, significantly contributing to more than 80% of the country's Falcata log production in recent years. Currently, Falcata plantations face several challenges, especially regarding plantation monitoring and management. The information on stand age is essential for the efficient monitoring and sustainable management of ITPs. With advances in technology and freely available satellite data, primarily from the Sentinel satellites, remote sensing has provided an alternative approach to determining stand age information. In this context, this study used multivariate regression models and machine learning classifiers, namely Support Vector Machine (SVM) and Random Forest (RF) to estimate Falcata stand ages from single-date and multitemporal Sentinel-1 and Sentinel-2 images and vegetation indices (VIs). The eleven multivariate regression models have R2 values ranging from 0.23 and 0.81 and with estimation errors of 1.72 to 3.58 years. The best multivariate regression model developed is an exponential model that relates Falcata stand age with the year 2021 Sentinel-2 surface reflectance bands and year 2021 Sentinel-1 VV and VH polarization bands. The model has training and validation data R2 of 0.61 and 0.55, which are the most consistent among the 11 regression models developed. When used for stand age estimation, this model may underestimate, or overestimate stand age by 1.72 years. When a machine learning approach is to be employed, the RF classifier performed better than SVM, particularly when estimating stand ages using multitemporal Sentinel-1 and Sentinel-2 data and VIs. The classifier has an overall accuracy of 84.69%, the highest among the eight classification results generated by the study.